Remove dead code, document functions, correct Rmd

[eliotmcintire]

Removes the commented-out code and an if (FALSE) {...} block in prepare_IgnitionAndEscapePredict(), adds short roxygen docs to every function, corrects the parameter/input/output desc strings, and fills in the Rmd (summary, events, links), which was still the template. Behaviour is unchanged: with comments ignored, every function parses identical to development except for the deleted if (FALSE) block, and parameter names, classes and defaults and input/output names and classes are identical. One metadata change to note: 12 inputs had their description passed positionally into sourceURL, leaving desc as NA; the text is now in desc and sourceURL is NA. tests/testthat passes. The .md/.html are left for the render workflow.

Removed functions: none (the only dead one, getRequiredFuelClasses, was already commented out).

Possibly unused, kept:

• parameters .plotInitialTime, .plotInterval, .saveInitialTime, .saveInterval, .useCache: never read.
• input fireSense_IgnitionFitted: only checked in .inputObjects, and only if whichModulesToPrepare has fireSense_IgnitionFit.
• input climateYear: only read in a check whose result (time) is never used; the layer taken is always year<time(sim)>.
• input propFlammable: set in .inputObjects, never read.
• argument timeStep of ageNonForest(): not used; TSD always increases by 1.

Seen, not changed:

• whichModulesToPrepare defaults to fireSense_EscapeFit, but the code tests for fireSense_EscapePredict.
• .inputObjects calls defineFlammable(rstLCC, ...); rstLCC does not exist when rstLCCs is supplied, and is a list otherwise.
• P(sim)$.studyAreaName, sim$studyArea and sim$rstLCCs are used but not declared in the metadata.

🤖 Generated with Claude Code

---

Tests (added in 3927403)

A testthat suite on a 4x4 toy landscape where every covariate is hand-computable. The toy climate encodes both the year and the cell in its values (year<Y> MDC cell n = (Y-2000)*100 + n, Tmax its negative), so asserting a value proves which layer of which variable was taken.

99 assertions, 0 failures, 0 skips. Verified to pass identically on origin/development and on this branch, run the way CI runs them (convertToPackage() + pkgload::load_all() + test_local()). Every substantive test was mutation-checked by breaking the module source and confirming the test fails: 14 mutations, 14 detected.

file	covers
test-metadata.R	exact input/output/parameter name+class contract
test-ageNonForest.R	ageing, burn reset, NA=0, timeStep ignored
test-getCurrentClimate.R	layer selection by year and variable, both compareGeom guards, supplied-raster passthrough, reuse message
test-init.R	declared outputs produced, climateVariablesForFire expansion, landcoverDT construction, event scheduling per whichModulesToPrepare
test-inputObjects.R	default climateVariablesForFire; two defects below
test-covariates.R	row set, column order, climate layer/year, non-forest columns, youngAge, igAggFactor aggregation and igAggFactor = 1, missingLCCgroup, fuel-class pins

Defects recorded, not blessed

Each states the intended assertion wrapped in expect_failure() -- so fixing the defect trips the test and forces the assertion to be unwrapped -- plus an assertion pinning the current state. No test asserts that buggy behaviour is correct.

1. .inputObjects leaves rstLCC_RTM NULL. The suppliedElsewhere("rstLCCs", sim) branch (l.509-510) does tail(sim$rstLCCs, 1)[[1]], but rstLCCs is not a declared input, so it is absent from the simList and the result is NULL. A rstLCC_RTM passed to simInit() is overwritten by this. Init() (l.241) then errors subscript out of bounds on .compareRas(rtm, NULL) -- the module cannot complete init as shipped. The test helper restores rstLCC_RTM after simInit() so the rest of the suite can exercise the code past this point.
2. .inputObjects cannot build flammableRTM (l.546): defineFlammable(rstLCC, ...) uses a local created only in the other branch, and the function is not imported.
3. Covariate tables do not advance with simulation time. getCurrentClimate() (l.296) only builds currentClimateRasters when it is.null(). The event reschedules annually, but from year two the object is populated and the build is skipped, so climate is frozen at start(sim) -- while the year column does track time(sim). Running 2001->2003 yields a table labelled 2003 carrying 2001 climate.
4. climateYear does not override time(sim). The inner lapply formal currentYear = time(sim) (l.322) shadows the outer currentYear (l.309-313). climateYear only reaches the "beyond the projection" message.
5. lightningDays is NaN in the ignition table. The layer is selected correctly (sim$lightningMaps["lightningDays"], l.416, returns the right values -- checked directly); the loss is inside fireSenseUtils::mergePreparedCovs(). Not diagnosed further.
6. Init() destroys categorical landcover on realignment. l.241-245 uses postProcess(..., to = rasterToMatch), which defaults to bilinear resampling; averaging class codes 19/16/210/20 gives values like 186.125 and 43.91, which match no class, so makeLandcoverDT() returns an all-NA landcoverDT -- silently, with the right rows and column names. Fix is method = "near".

Also confirmed: with the shipped whichModulesToPrepare default (fireSense_EscapeFit), only getClimateRasters and ageNonForest are scheduled and no covariate table is produced at all.

Honest labelling

The fuel-class magnitudes in test-covariates.R are a regression pin on fireSenseUtils internals, labelled as such -- not hand-derived. (The previous draft claimed they were "pinned from origin/development"; that claim was false and has been removed. They are in fact identical on both branches, verified by running the suite against each.)

---

Update (cbace6d)

This replaces "Possibly unused, kept" and the whichModulesToPrepare note above.

Removed; stop setting these:

• parameters .plotInitialTime, .plotInterval, .saveInitialTime, .saveInterval
• input fireSense_IgnitionFitted

Kept: .useCache (SpaDES.core reads it), climateYear (#13 makes it work), propFlammable (its only line is inside the block #13 rewrites; remove it after #13).

whichModulesToPrepare now defaults to fireSense_SpreadPredict, fireSense_IgnitionPredict and fireSense_EscapePredict. These are the three names doEvent tests for.

The save event now emits a message and does nothing else. The module never schedules it.

The version is not bumped here because #13 changes the same line.